Provenance Chain
Each number, its source file, and the query that produced it. Public, computer vetted figures throughout.
Principle
No figure appears in this report unless it can be regenerated from a public file by a stated query. These are public, computer vetted figures — none carried over from secondary reporting, and no estimates presented as measurements.
Data vintage
The FEC, Voteview and congress-legislators bulk files behind this report were downloaded on . The DIME contribution file was retrieved and reduced to aggregates on .
The date matters for reproduction. FEC bulk files are restated as committees file amendments, so a copy downloaded later will not necessarily match these figures row for row — the query alone does not fix the result. Anyone checking a number here should record the download date alongside the query, and should expect small divergences from a fresh copy rather than treat them as errors in either version.
Primary sources
Screens applied
Two bulk sources arrive contaminated in ways that are invisible until totalled. Both raw files remain in the database unaltered; the screens are separate tables, every excluded row is kept with its reason, and each threshold is computed from FEC-verified records rather than chosen by eye. The screens are therefore themselves falsifiable.
Self-filed independent expenditures
Anyone may file an FEC Form 5 for any amount, and nobody checks it. Seven filings — from “The Court of Divine Justice”, “The Committee of 300”, “The Masonic Illuminati Eye” and “Republican Emo Girl”, one of them targeting a candidate named Walter White — account for $43,899,177,138 of a $49,679,000,000 raw total. A row is discarded only when it fails two independent tests together: its target has no financial summary on file with the FEC, and its amount exceeds the largest expenditure ever verified against a candidate who did file one ($30,033,271). Either test alone would have been wrong — the first by itself discards 1,185 legitimate small expenditures against minor candidates. Duplicate amended filings are collapsed to the latest by file number first.
Screened total: $3.155B across 49,146 rows, 70% of the independent FEC transaction-file anchor of $4.505B. No figure published on this site was affected: the per-state pages draw on the FEC transaction file, whose largest single expenditure is the genuine $30,033,271. The contaminated file had not been used for any headline number.
Committee transfers in the DIME contribution file
DIME records committee-to-committee transfers, joint-fundraising distributions and party money alongside individual contributions, so a naive sum counts the same dollar at each hop: $60.46B against an FEC candidate-receipts anchor of $6.85B. The contributor-type column was reduced away during aggregation, so transfers cannot be identified directly — but transaction size settles it in one direction. The 2024 individual limit is $3,300 per election, so any transaction at or below $2,799 is necessarily within individual limits and cannot be a committee transfer.
That is a testable claim, and it holds. Contributions above $2,800 are excluded from every dollar total and retained only for geography and shares; their mean transaction runs $40,777 to $56,949 depending on office, an order of magnitude above any lawful individual gift.
Corrupt recipient records
Three defects are flagged separately rather than merged into one verdict, and rows that could not be checked at all are labelled as such rather than counted as clean.
- Invalid state (81 recipients, $3.411B). Presidential and national
committees carry state
00or blank, so every contribution scored as out-of-state and their out-of-state share read a flat 100%. - Exceeds FEC receipts by more than 10× (60 recipients, $16.013B). brown, sam records $8.42B against $24.1M reported to the FEC, and sheehy, tim $6.05B against $30.0M. This page previously attributed that to corruption in the source file. That was wrong, and the correction is set out below.
- Impossible mean contribution (2 recipients, $0.017B). A candidate averaging $1M or more per contribution cannot exist under any limit — mercer, lee shows 26 records averaging $64M. Committees are deliberately exempt from this test: Fairshake's $3.46M mean across 105 transfers is entirely real, and a naive version of this rule would have deleted it.
- Unverifiable (4,932 recipients, $10.237B). No metadata row exists, so no test could be applied. Held apart from the failures above — “not checked” and “checked and wrong” are different claims.
7,766 recipients pass all three tests. Screening moves the Senate out-of-state share from 94.5% to 85.3% on the DIME basis; this is a different measure from the 71.8% median reported below, which is computed per member from the FEC itemised individual file, and it does not replace it.
Correction — the 10× cases were this project’s error, not DIME’s
An earlier version of this page described the 60 flagged recipients as “corrupt at source” and a defect in a widely-cited academic dataset. A bug report to that effect was drafted and was not sent. Every record belonging to those recipients was then extracted from the source file and examined, and the diagnosis came back the other way.
Taking brown, sam: of 675,742 records, 99.1% of the dollar total is
transaction type 15E — contributions earmarked through a conduit — and
77,290 of the large ones carry the contributor name winred. Earmarked
contributions are itemised twice by design, once as the individual’s gift and once as the
conduit’s batch remittance. That is how FEC Schedule A works, and DIME reproduces it
faithfully. Summing amount by recipient without excluding conduit rows counts every
dollar routed through WinRed or ActBlue twice.
The arithmetic settles it. Restricting to records at or below the $3,300 individual limit gives $22,639,010 against the $24,063,366 he reported to the FEC — 94.1%, with the remainder in lawful PAC and party money above the band. The file was right. The query was wrong.
Three of the four hypotheses tested were refuted outright: transaction identifiers are unique
(duplicate ratio exactly 1.00 across all 60 recipients), no identifier had absorbed several
committees (the two recipient.name values per recipient are capitalisation variants),
and the amount field is not corrupt. What remained was a misuse of the file by this project.
The screen itself is kept, because its effect was right even though its stated reason was not: those totals are not candidate receipts and must not be published as such. It is now labelled as a conduit double-count rather than as corruption. The band screen described above is independently vindicated by the same finding — it works precisely because conduit batch remittances sit above the individual limit and individual gifts sit below it.
One narrow question does survive for DIME’s maintainers, and it is a documentation
question rather than a defect: whether the codebook warns that summing amount by
bonica.rid double counts earmarked contributions. That is the only form in which
anything will be raised upstream.
Reconciliation
| Measure | Raw | Screened | Independent anchor | % of anchor |
|---|---|---|---|---|
| Independent expenditures | $49.679B | $3.155B | $4.505B | 70% |
| Senate contributions | $19.851B | $1.820B | $1.724B | 106% |
| Presidential contributions | $8.664B | $2.044B | $2.718B | 75% |
| House contributions | $5.559B | $1.969B | $2.404B | 82% |
Anchors are the FEC transaction file and the FEC candidate summaries — sources independent of
the ones being screened. The residual runs the expected direction in every office, because the
FEC totals also contain PAC and party money above $2,800 that the screened tier excludes.
Reproduce with screen-money.sql, audit with usv_ie_excluded and
usv_dime_recip_screen.screen_status.
Headline figures and their queries
| Figure | Value | Derivation |
|---|---|---|
| IE against candidates | $2,555,156,720 | sum(amount) where tx_type='24A' |
| IE for candidates | $1,945,739,929 | tx_type='24E' |
| Direct to candidates | $545,392,705 | tx_type='24K' |
| IE : direct ratio | 8.25:1 | (24A+24E+24F+24N) ÷ 24K |
| Super PACs registered | 2,575 | cmte_type in ('O','U') |
| Committees making IE | 1,119 | count(distinct cmte_id) on 24A/24E |
| Top 1% share of IE | 49.5% | Window function over ordered amounts |
| Committee-to-committee volume | $17.83B | sum(amount) on oth24 |
| Party unity (R / D) | 92.3% / 96.9% | Vote matches party majority per roll call |
| r(PAC share, unity) R / D | 0.195 / 0.267 | corr() within party |
| Individual contribution records processed | 58,208,756 | Line count of itcont.txt (10.28 GB) |
| Aggregate rows retained | 231,832 | Grouped by committee × state × amount band |
| Total individual contributions | $15.65B | sum(total_amount) |
| Median out-of-state share (all) | 45.3% | percentile_cont(0.5) over 519 members |
| Median out-of-state share (Senate) | 71.8% | n = 96 |
| Median out-of-state share (House) | 40.1% | n = 423 |
| Congressional spend, all states | $6.49B | Receipts + IE, usv_state_cost |
| National mean per adult 18+ | $24.41 | Total ÷ 265,926,503 |
| Highest / lowest state per adult | $357.43 / $0.66 | Montana / District of Columbia |
Reproduction
The database is ioce_votes, separate from the Baltic ownership database so that
neither can affect the other. Loaders live in /opt/ioce/votes/:
schema.sql, load-fec.sql, load-votes.sql,
load-summary.sql, compute-metrics.sql and
aggregate-indiv.sh. Every file downloaded is listed above with its public URL; no
credentials are required to reproduce any figure on this site.
Any figure here that cannot be reproduced from the listed sources is an error and should be reported as one. The report is designed so that checking it does not require trusting it.